Processors have several different ways to access data called addressing modes.
This sentence is about "processors" in general, and not about a specific type of processor.
I think this is too generalized because you will always find an exception for such suggestions. Indeed, if you look at modern processors, you will find more exceptions than processors that follow this rule.
Indeed, for "simple" processors, such as 6800 or 6502, the instruction itself has one addressing mode:
lda $3A
... for example, uses the "zero page" or "direct" addressing mode.
Other processors really definitely had two different addressing modes in the same instruction. For example, the instruction "move" 68000:
move.w ($123).w, (a3, $4567)
For x86 CPU it is even harder to say:
In 6800, a command that can be compared with mov al, bl is called tba (without arguments), and mov al, [0x123] is lda $123 .
So, you could argue that mov al, bl is an instruction without an argument (the addressing mode is implied - since the instruction is written as movblal without any operands to other CPUs), and mov al, [0x123] is an instruction with one address argument memory (absolute addressing mode - because the command is written as ldal 0x123 with one operand to other CPUs).
(The only instructions from the original 8086 that don't let you argue seem to be those with address modes m8, imm8 and m16, imm16 , such as mov word ptr [123], 567 or add byte ptr [123], 45 )
Of course, you can also claim that the mov instruction and that al and bl are two arguments to the mov al, bl command.
Thus, it depends on your argument if the command mov al, bl is an instruction with the addressing mode "implied" (= without operands) or "register for registration".